Thank you for the blog post recommendation. I appreciate the resource on choosing embedding models.
However, the issue I’m facing isn’t with selecting the embedding model itself. The problem lies in the console’s inability to preview the vectorizer, which results in significant discrepancies in the semantic search. The distances returned are quite poor, as shown in the images and code I provided. I’m uncertain if there’s a mistake in my implementation of the Bedrock code.
Could you provide any guidance or insight on this?